Anthropic Adds Safer Auto Mode to Claude Code
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic launched an “auto mode” for its Claude Code tool, enabling the AI to make permission‑level decisions on users’ behalf.
- Auto mode acts as a middle ground, flagging and blocking potentially risky actions—such as deleting files or sending sensitive data—before they run and prompting the model to retry or ask for user intervention.
- Claude Code can operate independently, which Anthropic notes can be dangerous if it executes malicious code or hidden instructions.
- Team plan users can currently access auto mode only as a research preview.
- Enterprise and API users will receive access to auto mode in “the coming days,” according to Anthropic.
Why it matters: Developers and enterprises gain a practical safety net that curtails autonomous AI mishaps—like accidental file deletions or data leaks—while still leveraging Claude Code’s autonomous capabilities, though Anthropic warns the experimental tool doesn’t eliminate risk entirely and must be used in isolated environments.
